- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper investigates how shortcuts affect a model's ability to learn genuinely useful features. The authors first propose FSO for models with separate causal and shortcut branches, providing additional training signals to the causal branch. They then introduce GAP-FSO for standard neural networks, which suppresses shortcuts by penalizing the difference in logit margins between aligned and conflicting samples
Critical strengths
-
The paper analyzes the training dynamics under squared loss and cross-entropy loss separately, showing that the shortcut problems under these two losses are not the same.
-
Beyond synthetic experiments, the authors also compare their method with group-robust approaches using ResNet, DistilBERT, and multiple vision and text datasets.
Critical weaknesses
-
GAP-FSO interprets the difference in average logit margins between the two groups as a measure of shortcut strength. However, this equality relies on the assumption that the two groups have the same causal contribution and that the shortcut contribution changes sign exactly. In real-world datasets, even when samples are grouped by class, the difficulty of samples in the two groups may differ. In this case, reducing the gap does not necessarily mean removing the shortcut; it may also suppress useful predictive signals.
-
The connection between the theory and the practical method is not fully explained. FSO promotes learning in the causal branch through additional gradients, while GAP-FSO, which is actually used for ResNet and DistilBERT, penalizes the logit gap between groups. According to the paper's own derivation under cross-entropy loss, the latter does not change the updates to the causal term. Therefore, the accuracy improvements on real-world datasets do not necessarily demonstrate that GAP-FSO alleviates the causal starvation described in the paper.
-
The current experiments mainly report final accuracy, with limited validation of the underlying mechanism. The paper does not directly show whether reducing the gap also reduces the model's reliance on shortcuts such as color and background. Comparisons with other inter-group constraints under the same experimental settings are also missing.
Questions for authors
Could the authors conduct an experiment where the main object in an image is kept fixed while only the color or background is changed, and compare how the model's predictions change before and after training? It would also be useful to report whether these prediction changes correlate with changes in the logit gap. This would provide a more direct test of whether GAP-FSO achieves the claimed shortcut correction than simply comparing final accuracy.
5 comments
Response to W1 (2/2): Counterfactual Sensitivity and Core Probes
To examine the reliance on the shortcut independently of difficulty, we use a counterfactual swap: we fix the core subject (e.g., the animal or core text content) and alter only the surface shortcut attribute (e.g., rendering the image dark or bright, or substituting identity terms). Because the core content is identical, sample difficulty is held fixed, and any resulting shift in the model's output can be attributed solely to the shortcut. Based on this swap, we introduce two metrics using the decision margin (logit difference for CE; scalar output for MSE) and label : the average signed margin change cf_gap , and the average absolute score change cf_abs (used for text). Values closer to 0 indicate weaker shortcut reliance, where 0 means no change in the model's output; unlike the non-negative cf_abs, cf_gap is signed and therefore captures both the magnitude and direction of shortcut reliance. Since both metrics inherit the model's output scale, we compare ERM and GAP-FSO within the same model rather than across different backbones.
Turning to the empirical results in Table 2, GAP-FSO drives toward zero while cf_gap shifts from 0.262 without asymmetry to −0.055 under moderate and −0.450 under strong asymmetry, whereas ERM remains high between 3.1 and 3.9. This negative shift reflects a mild over-correction of the original shortcut consistent with : when the conflicting group is intrinsically harder (), forcing implies . Consequently, when within-class difficulty is tied to group identity, the confidence gap alone cannot be interpreted strictly as shortcut reliance. Even so, GAP-FSO's worst-group accuracy continues to improve under stronger asymmetry, achieving even higher performance than when cf_gap is positive (Table 1). This is consistent with our earlier hypothesis that the induced difficulty asymmetry itself acts as an informative predictive signal, suggesting that the model may leverage sample difficulty to aid classification on harder groups despite the apparent over-correction.
Table 2: cf_gap on Dogs-vs-Cats under varying difficulty asymmetry (CE loss).
| Difficulty asymmetry | ERM | GAP-FSO |
|---|---|---|
| None (symmetric blur) | 3.854 ± 0.536 | 0.262 ± 0.046 |
| Moderate | 3.134 ± 0.260 | −0.055 ± 0.046 |
| Strong | 3.342 ± 0.486 | −0.450 ± 0.123 |
In addition, to verify whether the reduction in shortcut reliance comes at the cost of core features, we fit linear probes on the frozen penultimate features alongside counterfactual sensitivity evaluations (cf_gap, cf_abs). Table 3 summarizes the results across both datasets, where the counterfactual swaps demonstrate that GAP-FSO substantially reduces the model's reliance on the shortcut. Meanwhile, the probes address a different question (whether this reduction compromises core information) and indicate that it does not: the core probe, defined as the worst-group test accuracy of a linear read-out from the frozen features, remains at least as high as that of ERM. Therefore, GAP-FSO successfully mitigates shortcut reliance without compromising core representation quality, confirming that the performance gains stem from genuine robust feature learning rather than collateral damage to core task information.
Table 3: Evaluation of ERM and GAP-FSO on Dogs-vs-Cats and CivilComments (CE loss).
| Dataset (model) | Method | Worst-group acc. | cf_gap | cf_abs | Core probe |
|---|---|---|---|---|---|
| Dogs-vs-Cats (ResNet-50) | ERM | 0.967 ± 0.002 | 1.427 ± 0.096 | 1.866 ± 0.070 | 0.978 ± 0.004 |
| Dogs-vs-Cats (ResNet-50) | GAP-FSO | 0.989 ± 0.001 | 0.165 ± 0.023 | 0.263 ± 0.023 | 0.989 ± 0.001 |
| CivilComments (DistilBERT) | ERM | 0.457 ± 0.048 | – | 1.942 ± 0.088 | 0.682 ± 0.030 |
| CivilComments (DistilBERT) | GAP-FSO | 0.687 ± 0.062 | – | 0.641 ± 0.249 | 0.691 ± 0.004 |
Response to W2 (1/2): How GAP-FSO Connects to Causal Starvation
A: We believe that, under the experimental design of these benchmarks, evaluating the model on out-of-distribution (worst-group) data is itself a partial verification of what the model has learned: since the shortcut no longer correlates with the label at test time, a higher accuracy indicates that the model relies more on core features rather than on spurious ones. The effect of GAP-FSO on causal starvation is indirect: in the idealized model, the penalty drives the shortcut component to zero () instead of leaving the permanent offset of uncorrected CE, so that the shortcut no longer absorbs the shared error signal. Under this premise, a higher OOD accuracy accompanied by a lower reliance on the shortcut is consistent with GAP-FSO alleviating causal starvation indirectly.
While OOD accuracy provides an indirect signal, we also conducted counterfactual evaluations, which measure this reliance directly. At fixed difficulty, cf_gap and cf_abs are much smaller than those of ERM on every model family that we evaluated (Table 4): cf_gap falls from 1.427 to 0.165 on Dogs-vs-Cats (ResNet-50) and from 3.657 to 0.764 on Waterbirds, cf_abs falls from 1.942 to 0.641 on CivilComments (DistilBERT), CLIP ViT-B/32 and Qwen2.5-0.5B show the same direction, and on Colored-MNIST cf_gap falls from 8.033 to 0.661. In Table 5, the share of examples whose predicted label flips when only the shortcut is swapped (flip rate) falls as well on the image tasks, for example from 11.9% to 4.4% on Waterbirds (ResNet-50), and changes from 13.2% to 12.4% on CivilComments with DistilBERT. More results can be seen in the table.
Table 4: Sensitivity of the prediction to the shortcut under the counterfactual swap (CE loss).
| Model and dataset | Swap | Metric | ERM | GAP-FSO |
|---|---|---|---|---|
| CNN, Colored-MNIST | colour of the digit | cf_gap | 8.033 ± 0.789 | 0.661 ± 0.177 |
| ResNet-50, Dogs-vs-Cats | dark / bright | cf_gap | 1.427 ± 0.096 | 0.165 ± 0.023 |
| ResNet-50, Waterbirds | water / land | cf_gap | 3.657 ± 0.117 | 0.764 ± 0.078 |
| DistilBERT, CivilComments | identity term | cf_abs | 1.942 ± 0.088 | 0.641 ± 0.249 |
| CLIP ViT-B/32, Dogs-vs-Cats | dark / bright | cf_gap | 3.132 ± 0.193 | 0.145 ± 0.202 |
| CLIP ViT-B/32, Waterbirds | water / land | cf_gap | 5.485 ± 1.570 | 0.339 ± 0.235 |
| Qwen2.5-0.5B, CivilComments | identity term | cf_abs | 3.583 ± 1.145 | 0.186 ± 0.102 |
Table 5: Flip rate under the counterfactual swap (share of test examples whose predicted label changes when only the shortcut is swapped; CE loss).
| Model and dataset | Swap | ERM | GAP-FSO |
|---|---|---|---|
| ResNet-50, Dogs-vs-Cats | dark / bright | 0.028 ± 0.001 | 0.005 ± 0.001 |
| ResNet-50, Waterbirds | water / land | 0.119 ± 0.014 | 0.044 ± 0.009 |
| DistilBERT, CivilComments | identity term | 0.132 ± 0.009 | 0.124 ± 0.017 |
| CLIP ViT-B/32, Dogs-vs-Cats | dark / bright | 0.047 ± 0.007 | 0.023 ± 0.007 |
| CLIP ViT-B/32, Waterbirds | water / land | 0.224 ± 0.046 | 0.069 ± 0.009 |
| Qwen2.5-0.5B, CivilComments | identity term | 0.096 ± 0.015 | 0.072 ± 0.012 |
Response to W1 (1/2): Confidence Gap under Difficulty Asymmetry
A: We thank the reviewer for this insightful question. The equality Δ(t)=2uf(t) indeed holds only under the theoretical idealizations of our framework (Proposition 5.1), where the causal contributions of the two groups are equal and cancel out under Assumption 2.1 (A1). On real-world data, this generalizes to Δ=2uf+δc, where δc denotes the difficulty gap between the groups, and the confidence gap alone may not disentangle the two terms. Similar idealizations are common in the theoretical literature on shortcut learning [1–4], so we examine empirically how a difficulty gap affects our method. Regarding class-level difficulty, the class-conditional (J=2) form compares aligned and conflicting samples within each class, so differences in difficulty between classes cancel and only the difficulty within a class remains. We therefore conduct controlled experiments on within-class difficulty asymmetry, examining both downstream performance and internal representations.
To this end, we construct a controlled Dogs-vs-Cats benchmark by applying Gaussian blur with different probabilities to the aligned and conflicting groups (none, moderate, and strong asymmetry, i.e., blur with probabilities (qa,qc)=(0.5,0.5), (0.25,0.75) and (0.1,0.9) on the aligned and conflicting images), making the conflicting group progressively harder. On test data, the resulting difficulty gap δc, estimated via a shortcut-free reference model of the same architecture, accounts for roughly 1%, 14%, and 25% of the ERM gap under CE loss. As shown in Table 1, GAP-FSO consistently outperforms ERM in worst-group accuracy across all difficulty levels, demonstrating its effectiveness even under confounding difficulty asymmetries. Although the margin narrows as asymmetry grows because ERM itself improves under stronger blur, we hypothesize that because blur correlates with group identity, models can exploit difficulty itself as an additional predictive cue (e.g., relying on background darkness only when the image is sharp), aiding performance on the harder samples.
Table 1: Worst-group accuracy on Dogs-vs-Cats under varying difficulty asymmetry (CE loss).